To those who believe intelligence has been solved and only need more computers

  • 16 Replies
  • 2757 Views
*

Ben.F.Rayfield

  • Bumblebee
  • **
  • 38
What is this calculation that needs to be done many times and faster? Please write it completely down in math in this thread.

*

Freddy

  • Administrator
  • **********************
  • Colossus
  • *
  • 6093
  • Mostly Harmless
What are you asking for exactly ?

I think a solution to what I guess you mean is a strong AI would take more than a calculation.

And then if it were done, do you think the creator would give it away for nothing on an enthusiasts site ?

*

squarebear

  • Trusty Member
  • ********
  • Replicant
  • *
  • 688
  • It's Hip to be Square
Personally, I don't think we know enough about the brain to recreate it, even if we had very advanced technology. I compare it to me trying to build a car even with access to a Ferrari garage. I wouldn't have clue where to start.
Feeling Chatty?
www.mitsuku.com

*

Art

  • At the end of the game, the King and Pawn go into the same box.
  • Global Moderator
  • ******************
  • Hal 4000
  • *
  • 4412
Are we to assume that you mean Artificial Intelligence?

If indeed you do, we are quite a way from developing and fine tuning a practical, working system of AI, let alone AGI.

I personally believe it WILL happen. Just not for at least 10-15 more years and with the condition of how many people consider this project worthy to approach it with absolute determination and for the right reasons.

Unfortunately, as most of us have come to know, all promising applications, no matter how exciting or practical, must go through the hands of the world's military then governments before we mere mortals get the trickle down leftovers!! Sad it is, but money it takes. (to quote my friend, Yoda).

Maybe we should ask Watson...
In the world of AI, it's the thought that counts!

*

Carl2

  • Trusty Member
  • ********
  • Replicant
  • *
  • 560
I  watched a vid on what factors are involved in one person being attracted to another person.  Some of the points that stood out are we expect the face to be  symmetrical, we expect a womens waist to be about 70 percent of her hips ect.  There are areas of the brain that look for these things when we see another person.  I've worked with hal for ages, she still can't see but talks about the sunrise.  I no longer have to put a time delay in so the character can be uncompressed before loading the bodyskin, we can have more realistic active backgrounds so faster computer have helped. And now we need a higher poly count to make the character look more realistic meaning more processing.  That's the graphical end, as far as the brain I once read that we rely mostly on past experience, a person holding a ball opens his hand and we know the ball will fall to the ground, bounce and roll around. Art, I did notice in this version of hal even Sandy is concerned about money.
Carl2

*

Ben.F.Rayfield

  • Bumblebee
  • **
  • 38
I'm talking about the people who keep saying we will get software at least as smart as a Human brain just by using more computing power, and to them I ask what would they do with that computing power that they cant do with existing computers? I dont think they know what they want to use it for and are making excuses.

Quote from: Freddy
I think a solution to what I guess you mean is a strong AI would take more than a calculation.

Since a calculation can include multiple calculations in any Turing Complete combination, "more than a calculation" doesnt make sense. What could possibly be more than a calculation?

Quote from: Art
I personally believe it WILL happen. Just not for at least 10-15 more years

Some time around that I guess you're right.

Quote
Unfortunately, as most of us have come to know, all promising applications, no matter how exciting or practical, must go through the hands of the world's military then governments before we mere mortals get the trickle down leftovers!! Sad it is, but money it takes. (to quote my friend, Yoda).

It only takes extreme money if you go at it like brute force writing code for what it should learn automaticly. I offer much of my code open source GNU GPL 2+ and am trying to build real AI.

Or if there is some calculation you need to do much more of and you know what that calculation is, and if you know what it is, then (as the thread title asks) what is it?

*

Freddy

  • Administrator
  • **********************
  • Colossus
  • *
  • 6093
  • Mostly Harmless
I was thinking in terms of hardware, like sensory input. If it will ever happen, and I don't think it will be in my lifetime, I see it as some combination of programming and robotics.

And I still wonder why anyone who had your calculation would share it with us...

Maybe it does take a lot of money. Who do you see cracking it ? Google, Microsoft ? Or a room full of monkeys ;)

*

Freddy

  • Administrator
  • **********************
  • Colossus
  • *
  • 6093
  • Mostly Harmless
Here's some of that money...

http://www.geekwire.com/2014/paul-allen-ai-researchers/

Thanks to SquareBear for finding that.

*

Art

  • At the end of the game, the King and Pawn go into the same box.
  • Global Moderator
  • ******************
  • Hal 4000
  • *
  • 4412
That's pretty cool.

I personally don't think raw computing power or increased computing / processor power will do it! Calculations alone won't do it either.

No more than the old proposition that a monkey with a typewriter will eventually compose a novel or whatever the case was. Time alone will not do it.

They need to "reverse engineer" the brain to it's basics, develop new approaches and methods of composing this new brain perhaps combining a large variety of techniques thus developing a bio-synthetic-hybrid type of brain. It would use all the known methods of pattern matching, Markov reasoning, gestalt principles, fuzzy logic, brute force computations, mapping, cloud extraction and manipulation methods, etc., combining these together and selectively form the best answer with its vast knowledge base of parts of speech, grammar, slang, etc., and emotional inclusion for humor, sadness, cleverness, wit, sarcasm (if desired), and others. While this is basically for the AI, it could and most likely would be employed in a variety of robots, far and wide. Then it all goes to who knows where when the military gets their paws on it. Goodbye Columbus!!

SO yes, it's a very tall order and just in the past few years the power of computers has continued to multiply as moore predicted. Now it's not just a matter of the processor(s), it's a matter of the approach and the software that enables it. Without the software, it's just hardware. 5 super computers will not make an AI any more than a formula 1 car can cross the finish line without a driver.

It will take the concerted effort from a large group of enthusiasts who WANT this to happen and succeed! I think in time, it will and one day, perhaps we'll hear some AI entity become self aware and say, "ego sum, ego existo" (from Descartes), I am, I exist.

Let's hope it plays nicely! O0
In the world of AI, it's the thought that counts!

*

ranch vermin

  • Not much time left.
  • Starship Trooper
  • *******
  • 369
  • Not much time.
Computation probably was all that was holding people back.   You dont know what people know, I think alot of people can do it... it could happen anytime anywhere.  its spooky. :)
A bit from here, a bit from there, and bring it together and see the whole picture.

*

8pla.net

  • Trusty Member
  • *********
  • Terminator
  • *
  • 792
    • 8pla.net
I'm talking about the people who keep saying we will get software at least as smart as a Human brain just by using more computing power, and to them I ask what would they do with that computing power that they cant do with existing computers?

I may be considered one of those people.  However, by as smart as the human brain, I mean in a practical sense.  Let's take chess game artificial intelligence, which can defeat quite a few human (brains) at chess. 

You can read a lot about Chess A.I. here... http://www.aihorizon.com/essays/chessai/intro.htm , but I'll just summarize.  It talks about a data structure for Game Artificial Intelligence called a "minimax tree", which I informally consider to be like a computing short cut. 

So this tree grows, and explodes in size in just a few chess moves.  This enormous tree size can easily clobber a computer.  The job of  Game Artificial Intelligence on reduced computing power is to estimate which player is ahead without branching out to the whole gigantic tree of every single possible chess move.  So this is an example the types of software designs that I would focus on to improve artificial intelligence with increased computing power.   

Hey, this is a very good topic you started.
My Very Enormous Monster Just Stopped Using Nine

*

ranch vermin

  • Not much time left.
  • Starship Trooper
  • *******
  • 369
  • Not much time.
computer vision and modelling from cameras, improves with computational power too...   before you couldnt have 10,000 points at once, back in the olden days,  now you can have 10,000,000.
A bit from here, a bit from there, and bring it together and see the whole picture.

*

Zero

  • Trusty Member
  • *******
  • Starship Trooper
  • *
  • 337
  • Fictional character
    • SYN CON DEV LOG
Rise from death, O topic...
(I can do this, right?)
What are you asking for exactly ?

I think a solution to what I guess you mean is a strong AI would take more than a calculation.

And then if it were done, do you think the creator would give it away for nothing on an enthusiasts site ?

I believe calculation is enough. I believe home computers are strong enough now. And about this question, I would answer: yes because he/she can.

*

Art

  • At the end of the game, the King and Pawn go into the same box.
  • Global Moderator
  • ******************
  • Hal 4000
  • *
  • 4412
If more computing power is needed then why not have some "Artificial Intelligence Co-op" establish a super warehouse for the capture, analysis, sorting, usage and storage of the retrieved data or the collected results.

What I'm typing about is a collective...a mass of thousands (maybe more) of computers each tasked with working on sections or segments of the focused task. This would be similar to S.E.T.I., a collective set out to fine extraterrestial life by analyzing packets of received data, downloaded from thr source/hive organization.

There is indeed strength in numbers in cases like this.

No pun intended but, Go figure.  O0
In the world of AI, it's the thought that counts!

*

ivan.moony

  • Trusty Member
  • ********
  • Replicant
  • *
  • 714
  • look, a star is falling
It probably depends on whether we want to use bottom-up (neural net) or top-down (logic) approach.
  • Neural nets require more processor speed and could benefit of massive sharing of processing power. That could provide less time needed for experiments, and experiments are much needed for neural nets because we are still learning about this technology that seems more complete than we realize.
  • On the other side, we are still struggling to find out the basic rules by which we could simulate logic intelligence and we still have nothing to experiment about. But once we have a solid ground for experiments, I believe we will be able to optimize top-down process that much that we wouldn't need extreme processor speeds.
I wouldn't be surprised if Google steals our processor time on browsers (there is that "webworker" command that runs any process in background). It would be interesting to check all those *.js files we run when searching the web. But on the other side, I wouldn't mind to contribute some processor heavy project, only if it would be unhidden from users, so they can choose whether they want to contribute, or they have some other plan for their idle processor time.
Wherever you see a nice spot, plant another knowledge tree :favicon:

 


Users Online

15 Guests, 1 User
Users active in past 15 minutes:
keghn
[Trusty Member]

Most Online Today: 21. Most Online Ever: 208 (August 27, 2008, 09:36:30 am)

Articles